I’ve been trying to get PyTorch to work on my AMD GPU using the ROCm drivers. And I’ve had to compile PyTorch from source in order to accomplish this (the prebuilt PyTorch binaries with ROCm enabled only support Vega GPUs and I have an RX 470). There were no errors during the build process and the installation, but it seems something must’ve gone wrong because when I try to just run “import torch” in the python shell, it appears to start loading for a few seconds before returning this and exiting:
/lib/x86_64-linux-gnu/libsystemd.so.0' with/lib/x86_64-linux-gnu/librt.so.1’ for IFUNC symbol
clock_gettime' python3: Relink/lib/x86_64-linux-gnu/libudev.so.1’ with
/lib/x86_64-linux-gnu/librt.so.1' for IFUNC symbolclock_gettime’
Segmentation fault (core dumped)
I can’t find any information on this error online, so any help in getting this to work would be greatly appreciated.